Table Of ContentContrasting Colonist and Indigenous Impacts on Amazonian Forests
Flora Lu1, Clark Gray2, Richard E. Bilsborrow3, Carlos F. Mena4, Christine M. Erlien5, Jason
Bremner6, and Stephen J. Walsh5
1Latin American and Latino Studies Department, University of California, Santa Cruz, CA
95064-1077, USA
2Carolina Population Center, University of North Carolina, Chapel Hill, NC 27516-2524, USA
3Department of Biostatistics, University of North Carolina, Chapel Hill, NC 27599-7420, USA
4Colegio de Ciencias Biológicas y Ambientales, Universidad San Francisco de Quito, Ecuador
5Department of Geography, University of North Carolina, Chapel Hill, NC 27599-3220, USA
6Department of City and Regional Planning, University of North Carolina, Chapel Hill, NC
27599-3140, USA
Abstract
In the most common narrative of tropical deforestation, agricultural clearing by migrant colonists
is assigned much of the blame and the activities of indigenous peoples are assumed to be
ecologically sustainable. We test this hypothesis using regional-scale survey and spatial datasets
from colonist and indigenous territories in the Northern Ecuadorian Amazon. We compare
measures of land use derived from surveys and satellite imagery for colonist and indigenous
households, and discuss regression models of the influences on land use for each of these
populations. The results confirm that forest impacts by colonists are greater than those of
indigenous peoples in terms of area cleared, rates of deforestation, and measures of forest
fragmentation. Nevertheless, substantial variation in land use patterns exists among five
indigenous groups. The results indicate that stereotypes of rapacious colonists and sustainable
indigenous peoples should be set aside as part of a more nuanced understanding of frontier land
use.
Introduction
The destruction of Amazonian rainforests, estimated at 2,500,000 ha/year through the
1990s (Malhi et al. 2008), has important implications for biodiversity loss, global climate
change, and livelihood security. Human residents of Amazonian forests and the forest frontier
include both indigenous peoples and mestizo colonists, populations which vary greatly in access
to land, labor and capital, degree of market involvement, and land tenure regimes. In addition,
these groups are widely assumed to differ in their environmental stewardship and conservationist
behaviors, with colonists seen as seeking material gain through extensive land clearing for
commercial agriculture while indigenous peoples are thought to possess cultural norms and
values that promote sustainable methods of resource use and conservation (Stocks et al. 2007).
The generalizations of rapacious colonists and ecologically noble indigenous groups belie
significant variation and highlight the importance of studies to illuminate the diversity of
demographic, economic and land and resource use practices within and between these groups.
The debate on indigenous conservation and sustainable land use illustrates the polarized
characterizations of these rainforest residents, with alternate portrayals as either a solution or as a
major threat to conservation (see, for instance, the debates in Conservation Biology volumes 7, 8,
14, and 15). In the former portrayal, they have been depicted as static, isolated, “ecologically
noble” conservationists living in harmony with nature (Brosius 1997; Conklin and Graham 1995;
Redford 1990). More recently, however, some biologists have become disillusioned due to the
contrast between the real-world behaviors of indigenous peoples and the “ecologically noble
savage” ideal, perceiving indigenous resource use as yet another threat to ecological viability
especially in light of population increase, technological change and market incorporation (e.g.,
Terborgh 1999). Their call to a return to people-free parks has been deemed by Wilshusen et al.
(2002) as a resurgent “protection paradigm” in international biodiversity conservation. In
rebuttal, Schwartzman, Colchester and others (Schwartzman et al. 2000a,b; Colchester 2000)
assert that indigenous peoples are potent political actors and an essential component of the
constituencies that are necessary for the long-term conservation of tropical forests, as seen from
examples such as the Kayapo in Brazil (Peres & Zimmerman 2001; Zimmerman et al. 2001;
Schwartzman & Zimmerman 2005). The framing of Native Amazonians in these debates as
homogeneous in terms of cultural values and use of land and resources has potentially negative
2
ramifications for developing more nuanced policies and long-lasting collaborations between
indigenous and non-indigenous stakeholders (Anderson 2001).
Views of colonists have tended to focus on their detrimental ecological impact on
rainforests due to land use strategies that, in their adaptation to former origin areas, are
inappropriate for Amazonian ecosystems and/or devised for income accumulation in a capitalist
economy. Research conducted among colonists in the Ecuadorian Amazon has highlighted the
importance of cash cropping, cattle raising, and wage labor to the household economy, and land
degradation occurs not only through intensive use (e.g., clearing and chemical inputs) but also
fragmentation and subdivisions of plots to multiple users (cite Dick’s work here). As proximate
agents of tropical deforestation, it is critical to understand the patterns, decision making, and
impacts of colonist land use, as demographic growth of migrants and their families in frontier
areas like the Amazon will continue to have a significant impact on ecosystem viability.
Understanding these processes among indigenous peoples is just as important, as these
populations are undergoing rapid processes of cultural, economic, and social change (citations)
and also control substantial areas of Amazonian rainforests (Peres 1994; Schwartzman &
Zimmerman 2005; Nepstad et al. 2006). However, in the literature there has been a relative lack
of studies comparing indigenous and colonist land use in the same region (exceptions include
Garland 1995; Godoy et al. 1997, 1998; Rudel et al. 2002; Hvalkof 2006; Stocks et al. 2007).
In this paper we report findings based on fairly large and representative samples of
indigenous peoples and colonists occupying the same forest frontier region in the Northern
Ecuadorian Amazon (NEA). The NEA is a particularly important study area because of its high
biological and cultural diversity, both threatened by ongoing rapid deforestation and forest
degradation. Since 1990 Ecuador has had the highest rate of deforestation in South America
(FAO 2005) due to rapid agricultural expansion, urbanization, land use intensification, and
petroleum exploitation. To compare colonist and indigenous land use in the NEA, we draw upon
parallel household surveys (Vadez et al. 2003) of the two populations conducted in 1999 and
2001.1 The 1999 survey of colonists involved return visits to farm plots originally visited in a
1990 survey, and collected data from 778 agricultural households in 64 colonization sectors or
areas (Bilsborrow et al. 2004). For indigenous households, surveys implemented with both
household heads and spouses were carried out among 499 households from 36 communities in
2001 (Holt et al. 2004). Unlike previous studies that of indigenous land and resource which are
3
commonly small-scale intensive studies, typically of one community (e.g. Santos et al. 1997;
Perrault 2005), our research encompasses five indigenous populations in the NEA study region:
the Huaorani, Kichwa, Cofán, Secoya, and Shuar. These groups span a gamut of population size
and density, history of contact with outsiders and settlement, and linguistic affiliation (Holt et al.
2004).
The emerging field of Land Change Science (Gutman et al. 2004; Rindfuss et al. 2004),
provides a framework for understanding the dynamics of forest frontiers such as those in the
Amazon (Lambin et al. 2001; Rindfuss et al. 2004). In following this approach, we integrate
survey, statistical, and spatial approaches to understand the human dimensions of environmental
and land use/land cover change. We present descriptive analyses of household land use,
summarize findings of multivariate models of factors influencing deforestation, and then present
key results from spatial analyses drawing on remotely sensed data. We use a remote sensing
image time-series (1986, 1996, 2002) to link biophysical data with household surveys and to
capture spatial and temporal contexts, providing characterization of landscape states and
conditions at a host of space-time scales. While a number of studies have now utilized remotely
sensed imagery to quantify land cover and land cover change in the Amazon (e.g., Behrens et al.
1994; Sierra 1999; Rudel et al. 2002a; Schwartzman & Zimmerman 2005; Nepstad et al. 2006),
fewer have used survey and statistical methods to investigate the drivers of forest resource
(Pattanayak & Sills 2001; Escobal & Aldana 2003; Coomes et al. 2004; McSweeney 2004) and
agricultural land use (Godoy et al. 1997; Godoy et al. 1998; Rudel et al. 2002a). Decades of
work by the authors in the NEA with its diverse human populations provides the kind of
multifaceted, large-sample data sets increasingly required to understand the complexities of these
coupled natural and human systems (cite).
Background and Study Site
The NEA study area includes parts of the provinces of Sucumbios, Orellana, Napo, and
Pastaza and borders the Andean foothills to the west and the Colombian and Peruvian Amazons
to the north and east (Figure 1). The region’s lowland moist tropical forests are among the
world’s most biodiverse (Pitman et al. 2002) and are part of the Amazon tropical wilderness area
(Mittermeier et al. 2003). The annual temperature averages 25 degrees Celsius with extremes of
15 degrees and 38 degrees. The annual rainfall is 2425-3145 mm, with an average humidity of
4
88% (Herrera-MacBryde & Neill 1997). The warm, wet climate fosters a wealth of biodiversity.
For instance, in the 600,000 acre Yasuni National Park (Figure 1), a UNESCO World Biosphere
Reserve, scientists have identified more than 600 species of birds, 500 species of fish, and 120
species of mammals (Kimerling 1991: 33). Moreover, a detailed assessment of tree biodiversity
in 16 tropical sites around the world conducted by the Smithsonian Tropical Research Institute
concluded that the park contained 1104 species (of at least one cm dbh) in a 25 hectare area, the
most of any site studied in the world (Romoleroux et al. 1997). In the Cuyabeno Reserve (Figure
1), another important Amazonian conservation area within the study region, 313 species of trees
have been identified within a single hectare, and 500 species of birds and 100 species of
mammals have been reported.2
The discovery of significant oil reserves at Lago Agrio by a Texaco-Gulf consortium in
1967 set in motion processes of road construction and agricultural colonization from highland
and coastal Ecuador which were facilitated by government settlement policies and fundamentally
driven by oil extraction (Hiraoka & Yamamoto 1980; Brown & Sierra 1994; Pichón 1997).
These processes have transformed the central part of the study area into an agricultural and
urbanizing landscape inhabited primarily by mestizo colonists from Ecuador’s coast and Andes
regions (Pichón 1997; Bilsborrow et al. 2004). This zone of colonization is bordered to the east
by Yasuní National Park and the Cuyabeno Wildlife Reserve and to the south by the large
Huaorani territory (Figure 1), all of which have also experienced peripheral agricultural
colonization (Greenberg et al. 2005; Messina et al. 2006). Agricultural colonization and oil
extraction have also dramatically altered the lives of Amerindian populations of the NEA who
have inhabited the region for millennia.
The total indigenous population of the Ecuadorian Amazon is over one hundred and fifty
thousand (INEC 2003), which is thirty percent of the total regional population and roughly
equivalent to the indigenous population of the entire Brazilian Amazon (Kennedy & Perz 2000).
The five populations included in the study—the Kichwa, Shuar, Huaorani, Cofán, and Secoya—
vary in population size, linguistic affiliation, history of contact, and economic activities (Holt et
al. 2004). The lowland Kichwa are the most numerous, with an approximate population of thirty
thousand in the NEA (INEC 2003) who primarily occupy discontinuous communal territories
(comunas) (CODENPE 2004). This group emerged in the aftermath of the violence and
depopulation associated with the Spanish conquest, when the Andean language Kichwa was
5
adopted as a lingua franca in mission villages of mixed ethnicity (Macdonald 1981). Under the
pressure of colonist encroachment, some Kichwa communities have adopted colonist-style
production and tenure systems (Macdonald 1981), and others have resettled away from the
colonization front (Irvine 2000), including along the region’s major rivers. The Shuar are
members of the Jivaroan language group and native to the southern Ecuadorian Amazon and
adjacent areas of Peru (Rudel et al. 2002a). Numbering approximately two thousand individuals
in the NEA (INEC 2003), they are the second largest indigenous population in the Ecuadorian
Amazon as a whole. Rudel and collaborators (2002a, 2002b) have described how agricultural
colonization of the Shuar traditional territory in the province of Morona-Santiago has pressured
some Shuar communities to adopt cattle raising and cash cropping. These experiences also likely
encouraged households to migrate to the NEA, where they have arrived as agricultural colonists
and settled in communal territories in a manner similar to the Kichwa.
The study also included three smaller indigenous populations. The language of the A’i
people or Cofán is believed by some to be unique, while others group it with the Chibcha family
of Colombia (Califano & Gonzalo 1995; Cerón 1995). Many Cofán were displaced from their
ancestral lands in the northern NEA by the initiation of oil extraction, and around 1000 Cofán
now live in six settlements in three separate territories (Townsend et al. 2005), including areas
within the heart of the zone of the colonization and within the Cuyabeno Wildlife Reserve and
the Cofán Bermejo Ecological Reserve (to the west of the study area) (CODENPE 2004). The
Secoya and related Siona, descendents of a larger group referred to as the Encabellado, belong to
the Western Tucanoan linguistic family and number approximately 800 people along the
Aguarico River and its tributaries in the NEA and adjacent Peru (Vickers 1993). Six Secoya and
Siona communities in the NEA occupy a contiguous ethnic territory, including areas within the
Cuyabeno Wildlife Reserve (CODENPE 2004). Finally, the Huaorani, whose language is a
linguistic isolate, are the least assimilated of Ecuador’s indigenous peoples and were peacefully
contacted for the first time only in 1958, having previously repelled outsiders through threats and
acts of violence (Rival 2002; Ziegler-Otero 2004). They are estimated to number two thousand
persons in twenty-eight communities and occupy a large contiguous ethnic territory in the south
of the study area, including areas within Yasuní National Park (CODENPE 2004). Most
Huaorani communities are distant from roads and market, but many have interacted with oil
companies which are active in the Huaorani territory (Lu 1999; Ziegler-Otero 2004).
6
Methodology
Survey Data Collection
This section describes the parallel survey data collections among colonists in 1999 and
indigenous peoples in 2001. The 1999 colonist survey was the second round of a longitudinal
data collection initiated in 1990. Prior to the 1990 survey, a probability sample of colonist farms
in the NEA was selected through a two-stage cluster sample with a sampling fraction of 5.9%. A
map of colonization sectors and farms from the government land reform and colonization agency
provided the sampling frame. Of 480 farms sampled in 64 colonization sectors in 1990, 408 were
inhabited and 418 resident households were interviewed. When these same farms were visited a
second time in 1999 they were inhabited by 778 agricultural households on large plots and 111
urban households on small residential plots due to rapid subdivision of farm properties. The 778
agricultural households interviewed in 1999 constitute the colonist sample analyzed in this paper.
The response rate for sample households in 1999 was 93%.
In 2001 a parallel household survey was carried out in 36 indigenous communities
representing five ethnicities: Huaorani, Kichwa, Cofán, Secoya, and Shuar. Since no sampling
frame was available for indigenous communities, a judgment sample of communities was
selected in the NEA region from the five ethnicities to capture a range of exposure to
colonization and external markets. In each study community a list of households was prepared
and used as a sampling frame to randomly sample 22 households per community, with all
households included in smaller communities. Complete information was obtained from 499
households with a response rate of 89%; these constitute the indigenous household sample
analyzed in this paper.
In both the 1999 and 2001 surveys two structured questionnaires were implemented in
each sample household, one with the male head of household and another with the female head
or spouse. These interviews were respectively conducted by male and female interviewers, most
commonly in Spanish but when necessary in the appropriate indigenous language with the aid of
a local interpreter. In single-headed households or in the case of a prolonged absence of one head
both questionnaires were administered to the person available. The 1999 questionnaires were
partially modified for the 2001 survey in order to better capture indigenous livelihoods and land
uses, but the two sets of questionnaires contain many common elements. In both cases the male
head’s questionnaire covered land tenure and use, production and sale of crops and cattle, off-
7
farm employment, and receipt of technical assistance and credit among other topics. Questions
on land use collected self-reports of the land area in various uses, and intercropped areas were
divided among constituent uses based on proportional coverage. The female head’s questionnaire
included a household roster and also asked about out-migration from the household, household
assets, and other topics.
Spatial Data Collection
Imagery
A classified time series of Landsat TM images (1986, 1996, 2002; Path 9/Row 60) was
used in this research. The time series imagery was classified using a hybrid supervised-
unsupervised classification method (Messina & Walsh 2001). An unsupervised classification
was applied first; the spectral signatures generated were evaluated using transformed divergence.
The results from the initial unsupervised classification were evaluated, and any classes that
displayed confusion were subset and run through the unsupervised classification separately.
These new signatures were added to the original signature set; this augmented signature set was
used for supervised classification. Training data for the supervised classification were obtained
from fieldwork in the study area.
Land cover classes in these images include forest, pasture, crops, barren, urban, and
water. Radiometric correction was applied, as examination of landscape change over time
necessitates radiometric correction so that pixel values are comparable between images (Song et
al. 2001). The 5s (Tanre et al. 1990) absolute radiometric correction algorithm was applied to
the image time series after the images had been converted to top of the atmosphere (TOA)
reflectance values.
Ground data for the classification of the Landsat Imagery were collected between 1999–
2001 by the Ecuador Project. Validation data were collected in 2002. Ground data were collected
using Global Position System (GPS) data with differential correction. The overall fuzzy accuracy
for the 2002 classification was 75.2%; however, the users’ accuracy for the forest class used in
this study was 95%.
Pattern Metrics
The term “pattern metrics” refers to a group of indices that have been developed for
evaluation of categorical maps (McGarigal et al. 2002). Landscape pattern metrics focus on the
composition and configuration of the classes included in categorical maps and thus the spatial
8
and geometric properties of these maps. Pattern metrics are commonly defined at three levels,
the patch, class, and landscape. Patch-level metrics are defined for individual patches and
characterize their spatial character and context, while class-level metrics examine all the patches
of a particular type, producing an average or weighted-average value depending on whether large
patches contribute more heavily to the index. Landscape-level metrics are integrated over all the
class types over the extent of the data, producing an average or weighted average value. The
metrics chosen for use in this study were chosen to represent composition (PLAND, PD, LPI)
and configuration (COHESION, AI) while minimizing redundant information. Metrics were
thus chosen carefully, as many pattern metrics are correlated, since they are based on a small
number of measurable patch characteristics, including patch type proportion, area, edge, and
connectedness (Riitters 1995).
Deforestation Rates
Forest cover may be described simply as the area affected by change by calculating forest
area at two time points and determining the difference between the two measurements. An
additional way of describing change in forest area is by calculating the rate of change. Studies
examining deforestation rates include Sader and Joyce (1988), Dirzo and Garcia (1992), and
Ochoa-Gaona and Gonzales-Espinosa (2000). Equation 1 shows yearly deforestation rate as
calculated by Sader and Joyce (1988), where F is forest area at the beginning of the period, F is
1 2
the forest area at the end of the period, and N is the number of years in the reference period.
F −F
1 2
F
Percent per year = 1 x 100 (1)
N
This study uses the formula used by both Dirzo and Garcia (1992) and Ochoa-Gaona and
Gonzales-Espinosa (2000) that calculate yearly deforestation rate as presented in equation 2,
where A and A are the forested areas at the start and end of the period being evaluated, and t is
1 2
the number of years within the period.
A − A 1
r =1−(1− 1 2)t (2)
A
1
9
Findings
Descriptive Analysis of Survey Data
To compare colonist and indigenous land use, we derived the total agricultural area
managed by each household as well as three constituent measures: area in pasture (almost
exclusively for cattle); area in crops planted for market sale (“cash crops”), primarily coffee; and
area in other crops that might be used for subsistence or sale, such as corn, rice and plantains
(“other crops”). Table 1 presents descriptive statistics of these measures for the colonists and
indigenous groups combined and separately. We also performed non-parametric Wilcoxon rank-
sum tests to compare the distributions of each measure between the colonists and the combined
indigenous (Table 1). Despite substantial heterogeneity across the five indigenous groups, the
results show that colonist households clear significantly larger agricultural areas overall, for both
pasture and cash crops, but indigenous households manage larger areas in other crops. Colonist
areas in all uses combined, in cash crops, and in pasture are, respectively, 2.1, 1.8 and 5.5 times
as large on average as those of indigenous populations, but the area in annual crops is only half
(0.48 times) as large. For both colonists and indigenous populations, a majority of the
agricultural area is devoted to market-oriented uses, which include pasture and cash crops.
To put these results in context, Table 2 presents mean values by ethnicity for key aspects
of the sample populations related to land use, including the percentages of households owing
cattle, using modern agricultural inputs (fertilizer, herbicides or pesticides), selling crops, and
hiring agricultural laborers in the past year. Consistent with their much larger areas in cash crops
and pasture but smaller areas in other crops, colonist households were more likely to participate
in all of these activities, and to participate in agricultural markets. The Secoya are a partial
exception, with many owning cattle, a livelihood strategy which was originally promoted by
missionaries from the Summer Institute of Linguistics (Vickers 1993), but has more recently
been facilitated by state-led development programs in the 1970s and 1980s and by the current
capital influx from Occidental Exploration and Petroleum Company who operate a concession
block in Secoya territory (Valdivia 2005).
Previous explanations of such differences between colonists and indigenous peoples have
focused on ethnocultural factors, such as different worldviews and value systems, but our data
show that the two populations also differ in other ways that might explain differences in land
use. As shown in Table 2, indigenous peoples are much more isolated from markets, have less
10
Description:Human residents of Amazonian forests and the forest frontier include both Progress of the victims: Political ecology in the Peruvian Amazon. In A.